Businesses replace drives and systems on a schedule regardless of true wear on them - their risk vs cost metric is very different than ours.
So used business hardware is a great value for those of us who have a sunk cost in labor/expertise.
Most drives I’ve owned have run for 10+ years, while any business won’t run them past 5. I can take the risk of losing a drive, while it’s more important for a business to prevent losing a drive. Business disruption is a more costly event.
DietPi is such a game changer.
Currently running it in a VM just for PiHole, and the ease of installing anything is nice.
Yea, small business routers (think old Cisco boxes) use a 5A laptop-style power supply. I suspect this is what OP’s using, as I’ve never seen a consumer model with such power specs. Consumer stuff is built around the cheapest hardware they can get away with.
Thing is, I don’t doubt the Cisco box (or similar business units) may need that much juice at times. Put your hand over one that’s not connected but powered on and they get hot.
Stuff like that they really don’t care about power consumption during design, just capability.
10 years?
My friend, many of us have been warning about SaaS since long before the public knew it as Cloud.
So you’re saying that the current Minecraft servers are all running on desktop hardware and not on standard U-chassis servers in racks in data centers?
If anything I’d expect them to run virtualized on standard rack servers.
Something here doesn’t add up. I don’t buy that enterprise server hardware that can out-perform desktop hardware by orders of magnitude can’t do what you need - especially since that’s surely how the existing servers run.
If this single-threading performance is really an issue, run some kind of virtualization on a rack server, build a VM with the exact hardware config you need, and the host hardware will make it happen (which I’d bet is exactly how current servers are run, because then they can be moved/Load balanced dynamically without missing a beat).
I currently have both single-core, single-cpu VM’s and multi-core, multi-cpu VM’s running on the same host for test purposes. The VM has no idea it’s virtual hardware is actually run on a different multi-cpu/core host.
Check out xcp-ng - it’s arguably the best open source virtualization platform today. VMware server can definitely do this, if you have a license.
That only works from PC to PC (it was my first thought too).
Besides track changes in word (already mentioned), I save versions of my work docs by appending the date and time to the file name. This makes it easy to confirm of we’re all looking at the same doc.
Example: worddoc_2026-09-08_16-46
Time uses 24 hour clock, so I never guess AM/PM, and it makes sorting really easy.
Nice thing is this works for any document type.
You should see the Tizen version…
Though neither are made by the Jellyfin team and they’re both dealing with the constraints of the underlying dogshit hardware/software
I have a lot of stuff self-hosted.
Just no outside access. Well, that’s not quite true, I do have a Tailscale VPN for remote management, but only certain machines are allowed to join, joins require my authorization, and access to the account is via two-factor.
The big thing with external access is to never permit it directly through your local connection. Do it though a Virtual Private Server (VPS) - let it get hit with access attempts. And then only allow specific traffic from the VPS to a specific service in your network using a VPN of some sort.
Two approaches:
Network isolation/DNS blocking - put the TV on a VLAN and only permit specific ports through to a media server (whatever the TV will support). If you lack VLAN capability, run PiHole as your DNS server (it works best as DNS and DHCP, but that’s just my opinion for things I like, such as proper name resolution for all devices).
Second option: (your idea with a Pi) disconnect TV from network entirely, then run something like Kodi (or whatever media center you prefer) on an old laptop, mini-PC, whatever, and have that device connect via HDMI. HDMI can passthrough the navigation controls from the TV remote (not guaranteed, but it’s part of the standard). I looked into this a long time ago but the performance wasn’t there yet. Today I’d probably try any distro and run Kodi on it.
Even my POS Samsung passes through enough of the controls to navigate Kodi on a 2017 laptop, while retaining power, volume, settings, etc for the TV itself.
My current setup is a Jellyfin server and Jellyfin client on the TV. I run PiHole with a blocklist for all the Samsung BS which has had the side effect of making the TV menus faster as it doesn’t retrieve any “what’s new” info from Samsung. Its actually kind of fun to monitor the FNS log and see the domains the TV hits, click Deny and see what breaks in the TV. On average about 30% of all DNS requests on my network get denied, and nothing of value has been lost. That’s how much crap is being sent all the time.
Hope that gives you some ideas.
Ah, you luddites amuse me.
The horses left the building a long time ago, and here you just now noticed and decide to play Don Quixote.
I mean it’s a commercial product - if stability issues were a common thing, that would be well known and published.
It’s a mesh VPN so it generally has the same stability challenges as any other: Hamachi, Nord, Meshnet, Wireguard (which is the underlying protocol that Tailscale uses).
If the Tailscale client is running on the NAS itself, I’d consider that the first thing to troubleshoot.
NAS’s generally don’t have much extra processing power, and in my experience apps designed to run on them aren’t the most robust. In this case, the VM/dock client has to consume all TS traffic, perform the encrypt/decrypt, put it back on the net (even if it’s only with the virtual net) over to Borg, which then has to write to the NAS. The NAS hardware has to do all this, with whatever little bit of ram it came with, also while writing all these changes, managing stripes, etc.
While they advertise these things can run stuff like this, it just seems like an opportunity for performance issues when the processing power design was for a NAS/RAID.
I have an ancient NAS on my network that could probably run the TS client, I just don’t bother. TS runs on all desktops and servers (for general access), but connectivity to non-TS devices goes through the server running as a TS router. That can be run on any device running TS - it’s just a feature to enable and configure.
I can’t imagine not using a proper internal DNS with split anymore.
I’m a big fan of Pihole (currently running it on a DietPi VM).
Being able to manage DHCP and DNS in a single place makes for much better control than the crappy DHCP from consumer routers. And almost any real DNS/DHCP server combo is better than those things.
Just rebuilt my PiHole over the last couple days, and it’s currently blocking 45% of DNS requests with no ill effects on any system - it’s a lot of Android app nonsense and a smart TV.
I’ve lost the Samsung TV service on the TV, but I never used it anyway, and the TV is so much more responsive now.
Yea, smart TV… Bad news. It’s next on the list to get a Kodi box on it.
I’m not - that’s a used price.
I don’t have a problem with used - all my drives are used, and they typically last 5-10 years. I have some drives from 2010 that are still running.
New drives fail too, so relying on drives being new to prevent failures is foolish. The only defense against failures is duplication (which 3-2-1 backup falls under).
Whats just as important is to use a file system that can detect and repair data errors (such as ZFS).
LTO tapes are 18TB today, raw - 48TB compressed.
How much data you got?
At that size I can probably build 2 Proxmox boxes with drives and compete very well on the costs.
A single 14TB disk is about $270 today - that’s in the same realm as a tape.
But tape also requires a drive, and that gets pricey fast.
This is the answer.
I haven’t done the math, but even SMB’s don’t use tape today.
They may do a local replica and then some kind of local-hardware-based cloud backup solution, because cost and capability.
Yea, getng people to do something is a challenge, let alone connecting a calendar to your server.
That’s where we should be today, but people really don’t understand how this stuff works, and won’t even try to learn even if you explain it in simple terms.
I’m upvoting for your openness and directness about your approach, what you will and won’t do.
You built this for you and are sharing, for us to do with it as we will.
It’s an interesting idea for basic printing.
Someone else mentioned just using a calendar client that can connect to different calendars.
I use Fossify to connect to calendars through Davx, which includes my Google calendar (configured it through DavX).
This way all entries remain independent. And I can filter my calendar view based on source calendar, and all entries are color coded (automatically) by which calendar they’re on.
I know it’s not exactly what you were asking for, buy it’s one way to approach it.
I get mine from a link on diskprices.com
Usually lead to ebay or Amazon - I’m not particular.
They’re almost always used business drives. Oldest one is now 4 years running.
Edit: My major exclusion is SMR, since I’ll be using them in a NAS of some kind (hardware or Proxmox). Today I’m all SATA - if you can, go with SAS, drive costs are lower.